On the reliability of composite analysis: an example of wet summers in North China

The validity of composite analysis is seldom discussed despite the fact that it can yield conflicting results. Our study confirms its validity by adding a reliability analysis to the classical composite analysis. Based on the signals extracted from composite analysis, 10 of the 14 wet summers in North China (1951-2020) can be "predicted" retrospectively. This study suggested the cyclonic anomaly over Mongolia at 500- and 850-hPa is closely associated with wet summers in North China. Interestingly, we found the most profound effects come from the Southern Hemisphere, with high confidence levels and large magnitude of the composite anomalies. These composite results are further cross-validated. We show that, based on the signals extracted from composite analysis, previously unseen wet summers in North China can be predicted with the mean absolute percentage error (MAPE) around 6%.